dsh-fs-encoding

DSH's built-in read, write, and edit tools only speak UTF-8: a GBK, Big5, or Shift-JIS file simply fails with invalid UTF-8 text, and a BOM-prefixed UTF-8 file loses its BOM on every save without so much as a warning. Three bytes gone from a file header is enough to break PHP, legacy compilers, or Windows software, and you will not discover the damage until something downstream goes wrong with no audit trail.

dsh-fs-encoding takes over those three tools while preserving every native safeguard (sandbox fencing, read-before-write gates, version checks). The core promise is byte-level fidelity: a file saved as GBK comes back as GBK, a BOM is restored exactly when it belonged and never injected when it did not, and CRLF / LF / CR line endings survive the round trip. Nineteen encodings are supported, from UTF-8/16/32 through GBK, Big5, Shift-JIS, EUC-KR, and the full Windows-125x family. If the target encoding cannot represent new content, the write is refused outright with a clear reason rather than silently filling the file with question marks. Three bonus tools insert, undo_last_edit, and str_replace_editor arrive with the same encoding governance.

It is aimed at anyone whose AI-assisted workflow touches CJK codebases, legacy Windows systems, or any non-UTF-8 file. The call signatures are identical to the originals, so no prompt rewrites or workflow changes are needed. Install it, keep using read / write / edit as before, and the encoding bookkeeping happens transparently underneath.